--- title: Dell and NVIDIA expand agentic AI infrastructure from desktops to data centers url: https://www.elseif.net/dell-and-nvidia-expand-agentic-ai-infrastructure-from-desktops-to-data-centers published: 2026-10-24T20:04:47+00:00 language: en section: Models source: https://www.aitimes.kr/news/articleView.html?idxno=42385 organizations: Dell Technologies, NVIDIA publisher: elseif --- # Dell and NVIDIA expand agentic AI infrastructure from desktops to data centers Dell Technologies and NVIDIA announced a collaboration to expand enterprise agentic AI infrastructure from desktops to data centers. The companies unveiled new hardware and software integrations that allow organizations to develop and operate AI agents within their own managed environments, supporting the transition from experimental projects to practical deployment. At an event on May 21 local time, Dell revealed an expanded portfolio of NVIDIA-based Dell AI Factory solutions and new desk-side agentic AI and AI-specific computing platforms. According to Dell, the number of customers using NVIDIA-based Dell AI Factory has grown to over 6,500, up from 5,000 announced in the previous quarter, indicating rising enterprise AI infrastructure demand moving toward actual adoption. Dell emphasized that expanding agentic AI requires infrastructure covering not only model performance but also inference cost, data control, security, and system operation. As AI agents repeatedly perform inference, tool use, and code execution, computing resource and token consumption increase, so enterprises need operational environments that can manage both performance and cost. Dell stated that using its own AI infrastructure allows organizations to strengthen control over data and operations compared to relying solely on public clouds. The company said its AI Factory can offer up to 90 percent higher cost efficiency compared to public clouds based on a two-year operation analysis, though actual savings may vary by workload and operating conditions. The portfolio expansion focuses on providing a unified system for needs from AI model deployment to actual service operation. By combining verified solutions and open software ecosystems, Dell enables enterprises to expand AI while maintaining security and management policies after proof-of-concept stages. Dell expanded its partner ecosystem to allow selection of AI models, applications, and operating systems. The company validated the Beam open-weight model from Reflection, and added Dynatrace AI observability, Mistral knowledge assistant and agentic AI platform, Fogsphere computer vision solutions, and SUSE Linux as operating system options. On security, Dell presented a multi-layered approach combining Red Hat AI Factory and NVIDIA OpenShell reference architecture. By linking hardware-based infrastructure, platform orchestration, and agent execution governance, the design integrates protection of runtime, network, data, and infrastructure layers. Through this, enterprises can operate AI agents in their own environments while managing execution rights and security policies, and receive support in reducing operational risks after deployment. Dell unveiled the NVIDIA-based Dell Deskside Agentic AI Solutions for NVIDIA, enabling developers and enterprises to build agentic AI in desk-side environments. Verified reference templates present the hardware, models, and software configurations needed for various use cases including agentic coding, video search and summary (VSS), NVIDIA NemoClaw-based agents, and AI practice environments for higher education. The solution is designed to scale from individual developer use to systems of over 1,000 enterprises, reducing the complexity of expanding verified AI environments across organizations. The hardware lineup is also strengthened. Dell Pro Precision with GB10 offers 64GB unified memory and can be clustered in configurations of two or four units to handle larger models and complex AI workflows. This configuration will be available starting October 23. Dell Pro Precision with GB300 is designed to provide up to 748GB of coherent memory and FP4 performance of 20 petaplops under Windows, supporting high-performance AI development and execution in existing desk-side environments managed by enterprises. At the data center level, Dell introduced new platforms tailored to AI agent workloads. The Dell PowerEdge M9882 configures 256 NVIDIA Vera CPUs in a single 48U rack, designed to work with GPU servers to handle agent orchestration, tool use, code execution, and evaluation. By separating CPU and GPU roles, the design reduces unnecessary resource input and improves power and cooling efficiency. The Dell PowerEdge R8822, aimed at general enterprise environments, is a air-cooled two-socket Vera CPU server installable in a standard 19-inch rack. It supports agent orchestration, sandboxing, data processing and analysis, and can accommodate up to four double-wide GPUs. The Dell PowerEdge XE9810, targeting large-scale inference markets, will feature 16 directly liquid-cooled computing trays based on the Grok 3 language processing unit (LPU). The product aims to provide high inference throughput needed for agentic AI services, complementing existing GPU clusters. Dell also supports AI utilization in communication networks. The NVIDIA Aerial RAN Computer platform running on Dell PowerEdge servers allows wireless access networks and AI workloads to share the same computing environment. This enables communication service providers to implement AI services at the network edge while simplifying infrastructure operation and reducing total cost of ownership. Additionally, Dell is preparing modular data centers based on the NVIDIA DSX POD reference design. By pre-integrating computing, storage, networking, rack-level power supply, and liquid cooling systems, the design reduces on-site installation and configuration work. Michael Dell, chairman and CEO of Dell Technologies, stated the companies are collaborating so enterprises can advance agentic AI from experimental stages to concrete results within infrastructure they own and manage. Jensen Huang, founder and CEO of NVIDIA, also emphasized that AI agents performing inference, tool use, and learning require new computing environments, and the two companies are building the foundations needed for the AI era from desktops to data centers. Product launches are scheduled sequentially from the fourth quarter of 2026 to the first quarter of 2027. The Dell PowerEdge R8822 is scheduled for October 2026 launch, while M9882 and XE9810 aim for first-quarter 2027 release. The modular data center based on the DSX POD reference design is expected to launch in early 2027.